How to Paste AI Markdown Tables Into Word and Google Docs Without Broken Borders

Paste AI Markdown tables into Word and Google Docs without broken borders

Copying a table out of an AI chat gives you two different things at once, and which one your paste uses decides whether you get a real table or fifty pipe characters. In my test on 2026-11-15, the chat app's Copy button followed by Ctrl+V in Google Docs produced a table object with 9 rows, 4 columns and 36 filled cells, while the same clipboard pasted as plain text produced no table at all and 50 pipes. The Word side needs one extra step: Pandoc wrote real Heading 1, 2 and 3 styles and a 36-cell table in 1.59 seconds, but the file it produced carries no cell grid lines and dropped every column alignment marker I gave it.

Key Takeaways

  • One Copy click puts two versions of the answer on the clipboard: an HTML version with real tags, and a plain-text version that is the raw markdown.
  • Normal paste into Google Docs kept 3 heading styles, a 9×4 table with 36 of 36 cells filled, bullet nesting at 2 levels, bold and monospace text, and left 0 stray markdown characters.
  • Paste without formatting of the same answer gave 34 flat paragraphs, 0 headings, 0 tables, 50 pipes, 2 bold markers and 8 backticks.
  • Turning on Tools → Preferences → Enable Markdown did not convert a text-only paste in my test, and the dedicated Paste from Markdown command asked for a Chrome extension before it would run.
  • Pandoc 3.9 -f gfm -t docx produced real Word heading styles, a hyperlink, a footnote and 36 of 36 cells, but no cell borders and 0 aligned cells.

How did I test these paste paths?

I sent one AI answer and one 536-word markdown report through four paste and export paths on 2026-11-15, then read the resulting document structure directly instead of judging by how it looked on screen.

Reading the structure matters because a paste can look fine in a screenshot and still be wrong underneath: a table that is really a paragraph of text with pipes, or a heading that is just bold text at a bigger size. For Google Docs I read the document through the Docs API and counted paragraphs by named style, table cells, bullets by nesting level and bold runs. For the .docx files I opened the file's XML and counted heading styles, table cells, borders, alignment properties, hyperlink relationships and footnotes.

Test conditions (run 2026-11-15, 14:00–14:50 KST)
ItemValue
Chat appGoogle Gemini web app, answer copied with its Copy button. Model name is not shown in that interface.
Document toolGoogle Docs in Chrome, read through the Google Docs API
ConverterPandoc 3.9 (-f gfm -t docx, and -f gfm -t html5)
Sample 1One Gemini answer: 1,566 characters of markdown, 231 words, 2 H2 headings, 1 H3 heading, one 4-column table with 8 data rows, nested bullets, one bold sentence, inline code, one python block
Sample 2A 536-word synthetic report I wrote: H1 1, H2 4, H3 1, a 4-column table with 8 data rows and alignment markers, nested bullets, a numbered list, a code block, one link, one footnote
Pass criteriaA table counted only if it existed as a table object with its own cells; a heading counted only if its paragraph carried a heading style; a link counted only if it carried a link relationship rather than a plain URL
Not measuredMicrosoft Word itself (no Word in this environment), ChatGPT and Claude copy buttons, Google Docs Paste from Markdown (blocked, see below), mobile apps

On a phone, swipe the table sideways to see every column.

What does the clipboard actually carry when you copy an AI answer?

One Copy click places two versions of the same answer on the clipboard, and a normal paste takes the HTML version while paste-without-formatting takes the plain-text one.

That is why the same chat answer behaves so differently in two attempts a minute apart. The chat app renders the markdown you see, and its Copy button copies both the rendered version (as HTML) and the source version (as text). The application you paste into then picks one.

The two versions on the clipboard after one Copy click
Clipboard formatWhat it holdsWhat a paste does with it
HTML (text/html)Rendered tags: heading elements, a table with rows and cells, bold and code spansGoogle Docs built 1 table, 3 heading paragraphs and 7 bullet paragraphs from it; 0 markdown characters arrived
Plain text (text/plain)The markdown source: hash marks, pipes, asterisks, backticks, one separator row of dashesGoogle Docs inserted 34 plain paragraphs; 50 pipes, 2 asterisks, 8 backticks and 3 hash-marked lines arrived as characters

On a phone, swipe the table sideways to see every column.

The plain-text version is not a degraded copy of the answer. It is the answer exactly as the model wrote it, which is what you want when the next stop is another tool. It is the wrong version when the next stop is a person reading a document.

How do the four paste paths compare?

The Copy button plus a normal paste is the only path that gave a real table with no cleanup, and Pandoc is the only path that produced Word heading styles, a working hyperlink and a real footnote.

The same AI answer through four paths (tested 2026-11-15)
PathWhat arrivedWhat did not
Copy button, then Ctrl+V in Google Docs3 heading paragraphs (2 at level 2, 1 at level 3), one 9×4 table with 36 of 36 cells filled, 7 bullet paragraphs at 2 levels, bold and monospace textThe code block: its fence and language label arrived as a plain "Python" line, and the code as monospace paragraphs
Copy button, then Ctrl+Shift+V in Google DocsAll 34 paragraphs of text, every word intact, 1,566 characters0 headings, 0 tables, 0 bullet objects, 0 bold runs; 50 pipes and 2 asterisks stayed in the text
Google Docs Paste from MarkdownNot measured: the command opened a dialog asking to install a Chrome extensionUnknown from this test: alignment, footnotes and code blocks on that path
Pandoc GFM to DOCXHeading 1 ×1, Heading 2 ×4, Heading 3 ×1 as real styles, 9×4 table with 36 of 36 cells, 7 bold runs, 1 hyperlink, 1 footnote, 12 list paragraphs at 2 levelsCell grid lines (the table style defines one rule under the header row), 0 aligned cells

On a phone, swipe the table sideways to see every column.

What happened when I pasted AI output with the Copy button?

A normal paste turned the answer into a styled document with no markdown characters left, and the table arrived with every cell filled.

I asked the chat app for a short report with headings, an 8-row spend table, nested bullets, bold, inline code and a python block, then clicked its Copy button and pressed Ctrl+V into an empty Google Document. These are the counts the Docs API returned afterwards.

What the AI answer became after one Ctrl+V (2026-11-15)
Element in the answerWhat the document held
2 H2 headings, 1 H3 heading3 paragraphs with real heading styles: 2 at heading level 2, 1 at heading level 3
4-column table, header row plus 8 data rows1 table object, 9 rows, 4 columns, 36 cells, 36 filled
Bullet list with 2 nested items7 bulleted paragraphs: 3 at level 0, 4 at level 1
Bold sentence, inline code4 bold runs, 4 monospace runs
python block, 2 lines4 monospace paragraphs, and a plain "Python" line from the fence label
Markdown markup0 asterisks, 0 pipes, 0 hash marks, 0 backticks

On a phone, swipe the table sideways to see every column.

Two details are worth knowing before you trust this path for a client document. The heading levels came through as the levels the markdown asked for, so an outline built from this document works. The code block did not: Google Docs has a code block you can insert, but the pasted content did not become one. I got monospace paragraphs with the fence language printed above them, and two private-use glyphs that render as small icons in the copy. If the document's value depends on a code block, plan to rebuild that one piece.

The paste also left 17 body paragraphs where the answer had fewer blocks, because empty paragraphs came in alongside the content. That is cosmetic on screen and annoying if you later count paragraphs to check a document.

What happens when you paste the same answer as plain text?

Paste without formatting keeps every character of the answer and none of its structure, and the count tells you exactly how much work you just gave yourself.

With the same clipboard in hand, Ctrl+Shift+V into a fresh Google Document produced 34 plain paragraphs, 0 headings, 0 tables, 0 bullet objects and 0 bold runs. The text was 1,566 characters, against 1,419 for the rich paste, and 73 of those characters were the markup itself: 50 pipes, 2 asterisks, 3 hash marks at line starts and 8 backticks. The table's dash separator row survived too, as a line of hyphens and colons.

  • You will see ## in front of every heading, because the heading is a character now.
  • Every table row is one paragraph of pipe-separated text, which is why the columns never line up.
  • Bold survives only as the asterisks around it, so the phrase looks louder than the sentence instead of being emphasised.
  • A fenced code block arrives with its opening fence, its language label and its closing fence as three ordinary lines.

Microsoft documents the same outcome for its own plain-text option: Keep Text Only discards formatting and non-text elements, and tables are converted to a series of paragraphs. That behaviour is described in Control the formatting when you paste text, which also lists the three paste options Word offers and how to change the default.

This is the moment to stop blaming the model. The asterisks in your Word document are not a hallucination; they are the markdown you asked the clipboard to carry as plain text.

Does turning on Markdown in Google Docs fix a plain paste?

Enabling Markdown did not convert a text-only paste in my test, and the dedicated paste command wanted a Chrome extension before it would run.

Google's help for using Markdown in Google Docs documents two different things, and they are easy to conflate. One is typing markdown as you write, where a line starting with hash marks becomes a heading. The other is converting markdown on paste, which the help page puts under a Paste from Markdown command. I tested the second one with the first one switched on.

  1. I opened Tools, then Preferences, and checked Enable Markdown. The checkbox confirmed as on.
  2. I pressed Ctrl+Shift+V with the same markdown clipboard: 34 plain paragraphs again, 50 pipes, 2 asterisks, 8 backticks. Identical to the run with Markdown off.
  3. I looked for Paste from Markdown under the Edit menu and under right-click. Both exist in the menu, and both opened a dialog titled Turn on copy, cut and paste, asking to install a Chrome extension and pointing to keyboard shortcuts instead.

So my result says more about this browser than about Google Docs. The command is in the product and Google documents it as a paste-time conversion; in a browser without that extension I could not run it. What I can say with a measured number is narrower: on this setup the setting alone did not change what a text-only paste produced.

Google also documents a route that needs no clipboard at all: open a markdown file from File, Open, Upload, and Docs converts it on import. I did not run that path in this test, so I am not putting a number next to it.

How do I get clean Word headings and a real table from Markdown?

Pandoc converts a markdown file to .docx in about 1.6 seconds and gives you real Word heading styles, a 36-cell table, a working hyperlink and a real footnote; what it does not give you is cell borders or column alignment.

For the DOCX side I used a second sample, a 536-word report with a 4-column table carrying alignment markers, a footnote and a link, so there would be something to lose. The command is one line.

pandoc -f gfm -t docx -o report.docx report.md
What the .docx held after Pandoc 3.9 converted a 536-word markdown report
In the markdownIn the .docx
1 H1, 4 H2, 1 H3Heading 1 ×1, Heading 2 ×4, Heading 3 ×1, as Word paragraph styles an outline can read
4-column table, header row plus 8 data rows9 rows × 4 columns, 36 cells, 36 filled, header row marked to repeat across page breaks
Alignment markers :--- and ---:0 cells carry an alignment property
Table bordersThe table uses Word's Table style; the only border that style defines is a bottom rule on the first row
7 bold spans7 bold runs
1 link1 hyperlink relationship, not a pasted URL
1 footnote1 footnote in a real footnote part, with its text
Nested bullets and a numbered list12 list paragraphs, 6 at level 0 and 6 at level 1
Fenced code blockCode paragraph style applied to the block
Markdown markup0 asterisks, 0 pipes, 0 hash marks, 0 backticks anywhere

On a phone, swipe the table sideways to see every column.

The alignment loss is not the markdown's fault. When I converted the same source to HTML first, all 36 cells carried their alignment as an inline style, 18 left and 18 right. The alignment disappeared in the DOCX step, in both the direct route and the HTML route I also ran. The border result is the same kind of fact: nothing in the file draws cell grid lines, so a table that looked boxed in the chat arrives as rows separated by one line under the header.

Both are fixable in a minute without touching the text. In Word, select the table and apply a table style or All Borders. Microsoft warns that table properties such as border style can arrive differently than they were in the source, in Copy and paste in Office for the web, which is the same warning from the other direction. If you convert the same documents often, set the look once in a reference document and pass it with --reference-doc; the flag is documented in the Pandoc manual. I did not test that flag in this session, so treat the reference-document advice as documented rather than measured.

The bigger win is that headings, footnotes and links arrive as objects rather than characters. A footnote that survives conversion is rare enough that it changed which path I use for reports with sources in them; the same trick appears in the workflow behind cleaning a messy CSV with AI without breaking totals, where the value was in what could be re-checked rather than what looked tidy.

Which path should you use for which document?

Choose by what the next reader needs to do with the file, not by which paste is easiest in the moment.

Choosing a path by the result you need
If you needUseWhy, from the numbers
A table that looks right in Google Docs nowCopy button, then Ctrl+V36 of 36 cells and 3 heading styles arrived in one paste, with 0 markdown characters left
The AI's exact markdown, for another toolCopy button, then Ctrl+Shift+V, or Copy as markdownThe source arrived unchanged: 1,566 characters including the 50 pipes a parser needs
A Word file with outline levels, a link and a footnotepandoc -f gfm -t docxHeading 1–3 as real styles, 1 hyperlink, 1 footnote, 36 of 36 cells
A Word file you will not touch afterwardsPandoc, plus a table style applied onceBorders and alignment are the two things the default output lacks

On a phone, swipe the table sideways to see every column.

If you write for a living, the pattern repeats often enough that it is worth memorising: rich paste for people, plain paste for machines, a converter when the destination is a file. The same split shows up in the six-step learning method, where the point of writing a concept down was that it could be read again by a human later.

Where does each path break down?

Each path fails in a different place: the plain paste fails on structure, the Copy button fails on code blocks, and the DOCX route fails on how the table looks.

  1. Plain paste fails on everything structural. Headings, tables, bullets and bold are characters, so fixing them is manual formatting, not a setting.
  2. The Copy button fails on code blocks. The fence and its language label came through as a "Python" line and the code as monospace paragraphs; Docs has no code block there.
  3. The Copy button also fails on anything the HTML does not carry. If the answer includes an image, a formula or a wide table the chat app scrolls sideways, check what arrived before you send the file.
  4. Google Docs Paste from Markdown was blocked here by an extension prompt, so I have no measurement for alignment, footnotes or code blocks on that path.
  5. Pandoc fails on the look of tables: no cell grid lines, and column alignment dropped in both the direct and the HTML route.
  6. None of these numbers describe Microsoft Word itself. Word was not available in this environment, so the Word-side statements above rest on Microsoft's paste documentation rather than on my own test.

What is your first step today?

Paste the same AI answer twice, once with Ctrl+V and once with Ctrl+Shift+V, and keep reading only the version that arrived with a table.

  1. Ask the assistant for the response, then click its Copy button once. Do not select the text by hand, because a hand selection is more likely to give you the text-only version.
  2. Paste into an empty document with Ctrl+V, then check that the table has cell borders and that the headings appear in the document outline.
  3. If the answer is going to a file rather than a person, save the markdown and convert it with pandoc -f gfm -t docx instead of pasting.
  4. Open the converted file once and count the table's rows against the original before you send it anywhere.

This guide is one of the nine mapped in AI Office Workflows That Actually Work: Verified Guides for Excel, PDFs, and Documents (2026) — the hub that puts each guide next to the number it measured, from the spreadsheet checksum to the four paste paths above.

Frequently Asked Questions

Why do I get asterisks and pipes when I paste from an AI chat into Word?

Because the paste used the clipboard's plain-text version, which is the raw markdown the model wrote. In my test the same clipboard pasted as plain text produced 50 pipes, 2 asterisks and 8 backticks with 0 tables. Pasting with formatting intact, or converting the markdown to .docx, avoids it.

Does Ctrl+Shift+V paste plain text in both Word and Google Docs?

It does in Google Docs, where my text-only paste arrived as 34 unformatted paragraphs, and Microsoft lists Ctrl+Shift+V as the Paste Text Only shortcut in Word. Microsoft describes that option as discarding formatting and converting tables into a series of paragraphs, which is the same result I measured in Docs.

Why is the table in my .docx missing its borders?

Pandoc's default table style defines only one border, a bottom rule on the first row, so nothing in the file draws cell grid lines. Applying a table style or All Borders in Word adds them, and a reference document can carry the look for every future conversion.

Can I keep footnotes when I convert markdown to Word?

Yes, if the markdown uses footnote syntax and the converter understands it: my .docx contained one real footnote with its text, not an inline bracket. I measured that with Pandoc 3.9 on a GitHub-flavored markdown file, so a different markdown dialect is worth testing once before you rely on it.

Comments

Popular posts from this blog

How to Clean Messy Excel and CSV Data With AI Without Breaking Your Totals

REA for AI Agents: What It Can Prove About an App

TrendRadar AI: Start With a Smaller News Brief